camisp-308
R
Gb
Gr
B
Bayer format with R
/Gr and Gb/B in alternate lines
Horizontal distance between same colors is 2
Public Version
Camera ISP Functional Description
www.ti.com
When the input source is memory, the preview engine always operates in one-shot mode. After enabling
the preview engine and processing a frame, the enable bit is turned off and it is up to firmware to reenable
it to process the next frame from memory.
When the input source is the CCDC, the preview engine can be configured to operate in one-shot mode or
continuous mode (
[3] ONESHOT).
The SBL image data read port is shared between the preview and CSI1/CCP2B receiver module. When
the image is read from memory, the read port must be affected to the preview receiver module by writing 0
into the
[27] SBL_SHARED_RPORTA. Programmers must ensure that the CSI1/CCP2B
module does not use this port before switching to the preview module.
6.4.7.1.3 Camera ISP VPBE Preview Input Formatter/Averager
Preview-engine output is limited to 4096 pixels per horizontal line to support 12 Mpix, due to line
memory-width restrictions in the noise filter and CFA interpolation blocks. To support sensors that output
more than 4096 pixels per line, an averager is incorporated to downsample by factors of 1 (no averaging),
2, 4, or 8 in the horizontal direction (
[1:0] COUNT). The horizontal distance between two
consecutive pixels of the same color to be averaged is selectable from among 1, 2, 3, or 4 for both even
(
[3:2] EVENDIST) and odd (
[5:4] ODDDIST) lines. This must be configured to match
the input pattern type.
For example, a Bayer pattern has a horizontal distance of two pixels of the same color for both even and
odd lines.
Valid output of the input formatter/averager is either 8 or 10 bits wide.
shows the horizontal distances for different patterns.
Figure 6-85. Camera ISP VPBE Preview Horizontal Distances for Different Patterns
6.4.7.1.4 Camera ISP VPBE Preview Dark-Frame Write
The preview engine is capable of capturing and saving a dark frame to memory instead of performing
conventional processing steps (
[7] DRKFCAP).
This dark frame can later be subtracted from the RAW image data to eliminate the repeatable baseline
noise level in the frame.
Each input pixel is written as an 8-bit value; if the input pixel value is greater than 255, it is saturated to
255. If a dark pixel is greater than 255, it is more likely to be faulty, and can be corrected by the
faulty-pixel correction module in the CCDC. If properly corrected, the value must be less than 255 when it
reaches the preview engine.
The
and
registers must be used to indicate the output
address and line offset, respectively, of the dark-frame output in memory.
6.4.7.1.5 Camera ISP VPBE Preview Inverse A-Law
To save memory capacity and bandwidth, the CCDC includes an option to apply 10-bit to 8-bit A-Law
compression and to pack the sensor data to 1 byte per pixel. To process this data correctly, the inverse
A-Law block is provided to decompress the 8-bit nonlinear data back to 10-bit linear data if enabled
(
[5] INVALAW). Even if the inverse A-Law block is not enabled, but the input is still 8 bits
(
[4] WIDTH), the data is-left shifted by 2 to make it 10-bit data. If the input is 10 bits wide, no
operation is performed on the data.
1208
Camera Image Signal Processor
SWPU177N – December 2009 – Revised November 2010
Copyright © 2009–2010, Texas Instruments Incorporated